BLM helicase suppresses recombination at G-quadruplex motifs in transcribed genes
van Wietmarschen, Niek; Merzouk, Sarra; Halsema, Nancy; Spierings, Diana C J; Guryev,
Victor; Lansdorp, Peter M
Published in:
Nature Communications
DOI:
10.1038/s41467-017-02760-1
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2018
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
van Wietmarschen, N., Merzouk, S., Halsema, N., Spierings, D. C. J., Guryev, V., & Lansdorp, P. M.
(2018). BLM helicase suppresses recombination at G-quadruplex motifs in transcribed genes. Nature
Communications, 9(1), [271]. https://doi.org/10.1038/s41467-017-02760-1
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
BLM helicase suppresses recombination at
G-quadruplex motifs in transcribed genes
Niek van Wietmarschen
1
, Sarra Merzouk
1
, Nancy Halsema
1
, Diana C.J. Spierings
1
, Victor Guryev
1
& Peter M. Lansdorp
1,2,3
Bloom syndrome is a cancer predisposition disorder caused by mutations in the
BLM helicase
gene. Cells from persons with Bloom syndrome exhibit striking genomic instability
characterized by excessive sister chromatid exchange events (SCEs). We applied single-cell
DNA template strand sequencing (Strand-seq) to map the genomic locations of SCEs. Our
results show that in the absence of BLM, SCEs in human and murine cells do not
occur randomly throughout the genome but are strikingly enriched at coding regions,
speci
fically at sites of guanine quadruplex (G4) motifs in transcribed genes. We propose that
BLM protects against genome instability by suppressing recombination at sites of
G4 structures, particularly in transcribed regions of the genome.
DOI: 10.1038/s41467-017-02760-1
OPEN
1European Research Institute for the Biology of Ageing, University of Groningen, University Medical Center Groningen, Antonius Deusinglaan 1, 9713 AV
Groningen, The Netherlands.2Terry Fox Laboratory, British Columbia Cancer Agency, Vancouver, BC V5Z 1L3, Canada.3Department of Medical Genetics,
University of British Columbia, Vancouver, BC V6T 1Z4, Canada. Correspondence and requests for materials should be addressed to P.M.L. (email:plansdor@bccrc.ca)
123456789
B
loom syndrome (BS) is a rare genetic disorder caused by
mutations in the BLM gene, which encodes the BLM
helicase
1. Symptoms of the disease include short stature,
immunodeficiency, UV sensitivity, reduced fertility, and a strong
predisposition toward a wide range of cancers. Cells from
BS patients display marked genome instability, characterized by a
10-fold increase in the rate of sister chromatid exchange events
(SCEs) in cells from patients compared with healthy controls
2,3.
SCEs are a byproduct of double-strand breaks (DSBs) or
collapsed replication forks that are repaired via homologous
recombination (HR)
4,5. Although SCEs are typically
non-muta-genic, they are considered markers for genome fragility and
somatic mutation rates
6. BLM antagonizes SCE formation by
dissolving double Holliday junction structures during HR, along
with its partners TOPO3α, RMI1, and RMI2
7,8. BLM also
promotes regression of stalled replication forks, facilitating fork
restart and preventing fork collapse and the formation of
DSBs
9,10. BS cells display higher numbers of
γH2Ax foci
11,
indicating frequent activation of the DNA damage response in the
absence of BLM. It has also been reported that BS cells display
elevated levels of loss of heterozygosity (LOH), due to exchanges
between homologous chromosomes
12–14. Besides its ability to
regress replication forks and dissolve Holliday junctions, BLM has
been shown to bind and unwind guanine-quadruplex
(G-quad-ruplex, or G4) structures in vitro
15–17. G4 structures are stable
secondary DNA structures that form at guanine-rich DNA
motifs
18,19and are known barriers for replication fork
progression
20.
Although SCEs can be used as a surrogate marker for collapsed
forks and DSBs, their locations could until recently only be
mapped cytogenetically at megabase resolution
21. This approach
does not allow investigations of the location and potential causes
of fork stalling and recombination in BS. We recently described a
single-cell sequencing-based technique, Strand-seq, which can be
used to map SCEs at kilobase resolution, enabling novel studies of
their locations and potential causes
22,23. Strand-seq is a single-cell
sequencing technique that relies on selective retention and
sequencing of DNA template strands after DNA replication and
cell division has occurred (Supplementary Fig.
1
a). SCEs are
detected as changes in orientation of DNA template strands
inherited by daughter cells. By sequencing DNA template strands
in single cells, changes in their directionality are identified and
mapped to the genome at kilobase resolution (Supplementary
Fig.
1
a, b).
Here we show that SCEs in BLM-deficient cells occur
fre-quently at sites of G4 motifs, especially those present in
tran-scribed genes. Furthermore, we show that although LOH events
appear to be more frequent in BLM-deficient cells, these events
were exceedingly rare in our study. We propose that besides
LOH, recombination at G4 motifs in transcribed genes is a major
contributor to genome instability and cancer predisposition in BS.
Results
Mapping of SCEs using Strand-seq. To address the question of
whether SCEs occur at random or at specific locations in the
genome, we performed Strand-seq on a panel of eight different
cell lines, four obtained from healthy donors (two primary
fibroblast and two EBV transformed B-lymphocyte cell lines) and
four cell lines from BS patients (two
fibroblast and two B-cell
lines) (see Supplementary Table
1
). We confirmed that the BS cell
lines displayed ~ 10-fold elevated SCE rates compared with wild
type (WT) (Fig.
1
a–d). Current Strand-seq libraries cover on
average ~ 1–2% of the genome due to loss of DNA during
pre-paration of single-cell sequencing libraries and uneven coverage
further limits the resolution of SCE mapping. The median
resolution of individual SCE mapping was ~ 10 Kbp (Fig.
1
e and
Supplementary Fig.
1
b) and
> 95% of all SCE could be mapped to
regions smaller than 100 Kb (Supplementary Table
1
). These
resolutions are several orders of magnitude higher than the
megabase resolutions than can be achieved by conventional SCE
mapping using cytogenetics
21.
We detected strong correlations between chromosome size and
the number of SCEs on each chromosome (Fig.
1
f, g), as one
would expect if SCEs were randomly distributed on a global level.
However, we also detected higher than expected numbers of
overlapping SCE regions in multiple common fragile sites (CFSs),
e.g., FRA3B (Fig.
1
h) and FRA7B (Supplementary Fig.
1
c), in the
EBV-transformed cell lines. The absence of SCE hotspots in CFSs
in primary
fibroblasts (Table
1
) suggest that this phenotype is
intrinsic to EBV-transformed B-lymphocytes, perhaps as a result
of replication stress induced by viral transformation
24. This is
consistent with previous observations that SCEs frequently occur
in CFSs in cells undergoing replication stress, presumably due to
replication
fork
stalling
and
collapse
25.
Strikingly,
SCE
frequencies within CFS hotspots are remarkably similar for the
WT and BS cell lines (SCE were mapped to any given hotspot in
~ 2–9% of libraries), even though BS cells display 10-fold
higher global SCE rates (Table
1
). This suggests that BLM has a
minor role in the processing of stalled or collapsed replication
forks at CFSs.
BS SCEs are enriched in transcribed genes. We next investigated
the distribution of SCEs relative to specific genomic features of
interest (FOIs). We developed a custom algorithm that compares
SCE distributions with simulated random distributions in relation
to a given FOI (see Methods section). For each cell line, we
performed a permutation analysis to calculate the frequency of
actual SCE regions overlapping with an FOI and compared it
against the expected background frequency. This analysis yields
relative SCE enrichments for a given FOI and allows for statistical
assessment of the strength of the association.
We
first turned to transcribed genes, as transcriptional activity
is a known cause of genome instability and mutations through
transcription–replication collisions and the formation of
co-transcriptional R-loops
26,27. BLM unwinds R-loops and the
absence of BLM has been linked to genome instability at sites
of R-loops
28,29. To study a possible link between SCE locations
and transcriptional status we assessed the transcriptional activity
in each of our 8 cell lines using RNA-seq. Genes were divided into
two categories based on the number of fragments per kilobase of
processed transcript per million fragments mapped (FPKM)
values: transcribed (FPKM
> 1) and non-transcribed (FPKM < 1),
resulting in an average of 60% (~ 23,000) of all genes classified as
transcribed and 40% (~ 16,000) as non-transcribed. A significant
enrichment of SCE regions overlapping with gene bodies was
found in all BS cell lines, but in none of the WT cell lines
(Fig.
2
a). However, these enrichments were not affected by gene
activity (Supplementary Fig.
S2a, b
). The same results were seen
after subsampling SCE regions from each cell line with the lowest
number of SCEs (WT1), indicating the detected SCE enrichments
are not an analysis artifact caused by the higher numbers of BS
SCEs (Supplementary Fig.
2
c). SCEs were also significantly
enriched in the gene promoter region of BS cells, independent of
the transcriptional status of the associated genes (Supplementary
Fig.
2
d–f). We also investigated if gene expression levels affected
SCE occurrence within those genes. To do this, we divided all
expressed genes into four categories based on their RPKM-values,
ranging from low to high expression, and assessed the number of
SCEs overlapping the genes in each category. We found only
weak-to-moderate correlations between gene expression levels
and SCE occurrence in all eight cell lines (R
2-values ranging from
0.05 to 0.64), with no differences between the WT and BS cell
lines (Supplementary Fig.
2
g). These results indicate that
transcription by itself does not appear to have a strong role in
SCE formation.
BS SCEs are enriched at G4 motifs. We next considered the
possibility that the intragenic SCE enrichments might be caused
by the presence of G4 in and around genes. BLM is known to
bind and unwind G4 structures in vitro
15–17and G4 motifs occur
frequently within gene bodies and promoters
30,31. To assess SCE
chr1 chr2 chr3 chr4 chr5 chr6 chr7 chr8 chr9 chr10 chr11 chr12 chr13 chr14 chr15 chr16 chr17 chr18 chr19 chr20 chr21 chr22 chrX chrY BS fibroblast: 39 SCEs WT fibroblast: 4 SCEs
a
b
c
d
f
100 90 80 70 60 50 40 30 20 10 0 SCEs mapped (%) 80 60 40 20 0 75 60 45 30 15 0 WT1 WT2 BS1 BS2 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 5.0 4.0 3.0 2.0 1.0 0.0 Cell line 50 100 150 200 250 50 100 150 200 250 Chromosome size (Mbp)h
chr3 (p14.2) 500 kb 60,100,000 60,200,000 60,300,000 60,400,000 60,500,000 60,600,000 WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 FHIT 60,700,000 60,800,000 60,900,000 hg19 24 29 Chromosome size (Mbp) WT3 WT4 BS4 <10 bp <100 bp <1 Kbp <10 Kbp <100 Kbp <1 Mbp <10 Mbp BS3 Cell line Resolution p = 7.5e–189 p = 1.7e–243 WT1 WT2 WT3 WT4 WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 BS1 BS2 BS3 BS4 SCEs/library SCEs/library SCEs/chr/lib SCEs/chr/lib R2 = 0.92 R2 = 0.92 R2 = 0.90 R2 = 0.90 R2 = 0.94 R2 = 0.94 R2 = 0.95 R2 = 0.96e
g
enrichments at G4 motifs, we determined the distributions of the
canonical G4 motif (G
3+N
1–7G
3+N
1−7G
3+N
1−7G
3+) across the
genome using a custom algorithm, and performed our SCE
enrichment analysis on these regions. For this analysis, we used a
stringent 10 Kb size cutoff for SCE regions included in this
ana-lysis because G4 motifs occur frequently throughout the genome
(~ 8.6 Kb on average) and including larger SCE regions would
result in increased noise because of the high likelihood of
(per-mutated) SCE regions overlapping G4 motifs purely due to their
size. Strikingly, we found significant, ~ 20% enrichments over
expected levels for SCE regions overlapping G4 motifs in the BS
cells, but no enrichments in the WT cells (Fig.
2
b), indicating that
G4 structures are a causal factor for SCE formation in absence of
BLM. We subsequently tested if presence of G4 motifs in genes
had an effect on SCE enrichments by splitting all genes into those
with and those without G4 motifs. Although we detected
sig-nificant SCE enrichments in BS cells for both genes with and
without G4 motifs, these enrichments were stronger for genes
containing at least one G4 motif (Supplementary Fig.
3
a, b),
indicating that the presence of G4 motifs is at least partially
responsible for the SCE enrichments detected in genes in BS cells.
A similar result was seen for SCE overlapping promoters with or
without G4 motifs (Supplementary Fig.
3
c, d). Based on these
results, we decided to further investigate the link between G4
motifs, transcription, and SCE formation in BS cells.
We detected
> 350,000 canonical G4 motifs in the human
genome, consistent with previously reported numbers
32. However,
cells may harbor only ~ 10,000 actual G4 structures
33. As our
analysis is based on SCEs overlapping with G4 motifs, we likely
overestimate overlaps with G4 structures in our permutation
analysis, leading to reduced enrichment estimates. The high
prevalence of G4 motifs also means that larger SCE regions are
likely to overlap with at least one G4 motif purely by chance in our
permutation analysis, leading to elevated noise in our analysis and
reducing relative enrichment values of the observed SCE regions.
Using less stringent size cutoffs for SCE regions size than the 10 Kb
cutoff used for Fig.
2
b did indeed decrease the relative SCE
enrichment values for the BS cells, but not the WT cells
(Supplementary Fig.
4
a), although BS SCE enrichments remained
significant for all cutoffs used. Next, we added increasingly large
flanking regions to the observed SCE regions to increase random
overlaps, potentially decreasing SCE enrichment values. We did
indeed observe an inverse relationship between SCE enrichments
and the size of
flanking regions in BS, but not WT cells lines
(Supplementary Fig.
4
b). This result suggests that even at our
stringent 10Kbp cutoff for SCE region size, there is noise present in
the permutation analysis. Taken together, we conclude that actual
SCE enrichments at G4 structures are almost certainly much higher
than reported in this study. As including larger SCE regions only
affects SCE enrichments for BS cell lines, we also conclude that the
enrichments we detect are indeed specific for BS cells.
Besides the canonical G4 motif, we also tested alternative G4
motifs for SCE enrichments. We did detect BS SCE enrichments at
sites containing G4 motifs containing smaller (n1–3) and larger
(n1–12) spacer regions (Supplementary Fig.
4
b, c), although the
enrichments were not as strong as for the canonical motif. This
suggests that the canonical G4 motif is more likely to form
G4 structures and induce SCE formation in vivo, or that BLM
displays some specificity for G4 structures with medium-sized
loops. Significant SCE enrichments were also detected at previously
described
“observed quadruplex regions” (Supplementary Fig.
4
d),
reported to constitute all regions in the genome capable of forming
Fig. 1 High-resolution mapping of SCEs and common fragile site hotspots. a, b Representative Strand-seq libraries generated from a a WTfibroblast and b a BSfibroblast. Mapped DNA template strand reads are plotted on directional chromosome ideograms; reads mapping to the Crick (positive) strand of the reference genome are shown in green, those mapping to the Watson (negative) strand are shown in orange. SCEs are identified as a switch in template strand state, indicated by arrowheads.c, d Number of SCEs detected during a single cell cycle in c primaryfibroblasts and d EBV-transformed B-lymphocytes obtained from healthy donor and BS patients. Each grey point represents number of SCEs detected in a single-cell Strand-seq library, red lines indicate mean± SD. p-values were calculated using ANOVA. e SCE mapping resolutions across all eight cell lines. Lines represent percentage of the total number of SCEs mapped at resolutions below indicated values.f, g Correlations between average numbers of SCEs/chromosome/library and chromosome size forf WT and g BS cells.R2-values are color-matched to the cell lines.h Example of SCE hotspot detected within FRA3B (FHIT). Mapped SCE regions for each cell line were uploaded onto the UCSC Genome Browser. Black bars represent genomic locations of SCE regions; size indicates mapping resolution using the BAIT program. Red box indicates the location of the SCE hotspot as detected by Strand-seq
Table 1 SCE hotspots in common fragile sites, related to Fig.
1
Cell line CFS Coordinates SCE hotspot Genes(s) Libraries with SCE in hotspot
SCEs (n) p-value WT3 FRA3B chr3:60,235,449–60,601,404 FHIT 15/334 (4.5%) 2128 3.2e−18 FRA7B chr7:5,880,409–6,120,310 CCZ1, PMS2 11/334 (3.3%) 2128 4.8e−13 FRA7B chr7:6,707,575–6,967,206 PMS2CL, CCZ1B 29/334 (8.7%) 2128 8.9e−50 WT4 FRA3B chr3:60,235,449–60,601,404 FHIT 28/320 (8.8%) 2326 1.5e−42
FRA5H chr5:58,828,828–59,116,208 PDE4D 10/330 (4.0%) 2326 3.8e−10 FRA7J chr7:69,775,932–69,923,390 AUTS2 8/330 (2.4%) 2326 8.2e−9 FRA16D chr16:78,463,763–78,719,484 WWOX 8/330 (2.4%) 2326 3.6e−7 BS3 FRA1A chr1:10,628,012–10,771,196 PEX14, CASZ1 13/326 (4.0%) 14,667 4.2e−9 FRA3B chr3:60,235,449–60,601,404 FHIT 13/326 (4.0%) 14,667 1.3e−4 FRA10G chr10:52,320,213–52,640,424 ASAH2B, A1CF 14/326 (4.3%) 14,667 3.2e−6 FRA16D chr16:78,463,763–78,719,484 WWOX 13/326 (4.0%) 14,667 2.8e−6 BS4 FRA7B chr7:5,880,409–6,120,310 CCZ1, PMS2 11/334 (3.3%) 2128 4.8e−13
FRA7B chr7:6,707,575–6,967,206 PMS2CL, CCZ1B 11/306 (3.6%) 12,724 5.0e−5 FRA11G chr11:103,881,556–104,600,858 PDGFD 17/306 (5.6%) 12,724 2.4e−5 FRA17q21 chr17:43,517,064–43,840,133 PLEKHM1, LRRC37 10/306 (3.3%) 12,724 6.7e−3
Overview of all SCE hotspots detected in human WT and BS cell lines, as well as frequency of SCE occurrence within hotspots. SCE hotspots were only detected in EBV-transformed cell lines and all hotspots occurred within known CFSs.
quadruplex structures
34. As before, SCE enrichments were not
affected by SCE subsampling (Supplementary Fig.
4
e). We could
also exclude that enrichments were caused by nucleotide slippage or
high GC content, as SCEs were specifically depleted in genomic
regions with A-rich motifs (A
3+N
1−7A
3+N
1−7A
3+N
1−7A
3+) or high
GC content across all eight cell lines (Supplementary Fig.
4
f, g).
Taken together, these results support that G4 structures are a major
cause of SCE formation in BS cells.
BS SCEs map to G4 motifs on transcribed strands. As
tran-scription can promote the formation of G4 structures
18and G4s
were shown to occur mainly in euchromatic regions of the
gen-ome
33, we hypothesized that BS SCEs occur at G4 motifs in
transcribed genes. We therefore divided all genes into four
categories based on (1) whether genes are active or silent and (2)
the presence or absence of intragenic G4 motifs, and performed a
separate SCE enrichment analysis for each category. We detected
the strongest BS SCE enrichments in transcribed genes containing
at least one G4 motif, whereas non-transcribed genes lacking G4
motifs did not show any significant SCE enrichment patterns
(Fig.
2
c–f). This points to a synergistic effect of transcriptional
activity and the presence of G4 motifs in genes on the enrichment
of SCEs in BS cells.
Intragenic G4 motifs can occur either on the transcribed or
non-transcribed strand (Fig.
3
a) and it is believed that this G4
motif
‘strandedness’ affects how G4 structures influence gene
expression
35. To assess whether G4 strandedness affects SCE
formation, we separated all intragenic G4 motifs into different
categories based on strandedness and transcriptional status of the
Active genes with G4 motifs
Enrichment
Silent genes with G4 motifs
0.8 0.9 1.0 1.1 1.2 1.3 0.8 0.9 1.0 1.1 1.2 1.3 Enrichment
Silent genes without G4 motifs Cells Cells Cells Enrichment G4 motifs 0.8 0.9 1.0 1.1 1.2 1.3 WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells Enrichment Genes 0.9 1.0 1.1 1.2 Enrichment 0.8 1.0 1.2 0.5 1.0 1.5 2.0 Enrichment
Active genes without G4 motifs
a
b
d
f
0.663 0.334 0.546 0.912 <0.001<0.001 <0.001<0.001 0.042 0.288 0.338 0.191 <0.001 <0.001<0.001 <0.001 0.543 0.993 0.026 0.806 <0.001<0.001 <0.001<0.001 0.617 0.785 0.009 0.005 <0.001 0.014 0.037 0.048 0.889 0.403 0.437 0.773 0.09 0.056 0.004 0.017 0.811 0.931 0.998 0.999 0.074 0.016 0.855 0.003***
***
***
*
***
***
***
***
***
***
***
***
***
**
***
*
*
*
*
**
*
**
*
c
e
Fig. 2 Bloom syndrome SCEs are enriched at G4 motifs in active genes. Relative SCE enrichments (red points) over random distributions (violin plots) for SCEs that overlap with one or multiplea genes; b G4 motifs (G3+N1–7G3+N1-7G3+N1-7G3+);c active genes containing one or more G4 motifs; d active genes
without G4 motifs;e silent genes containing one or more G4 motifs; and f silent genes without G4 motifs. All values were normalized to the median permuted value for overlap of SCEs with FOIs (out of 1,000 permutations) and relative SCE enrichments over these values were plotted on the y-axis. p-values indicate the fraction of permuted overlaps (out of 1,000 permutations) equal to or higher than overlap with observed SCE regions. Significant p-values are indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001
G4 motifs non-transcribed strand silent genes 0.7 0.8 0.9 1.0 1.1 1.2 Enrichment Enrichment
G4 motifs transcribed strand silent genes
0.7 0.8 0.9 1.0 1.1 1.2 Enrichment
G4 motifs non-transcribed strand active genes
0.8 1.0 1.2
Enrichment
G4 motifs transcribed strand active genes
0.8 1.0 1.2
Enrichment
Enrichment
G4 motifs non-transcribed strands
0.8 0.9 1.0 1.1 1.2 Cells Enrichment
G4 motifs transcribed strands
0.9 1.0 1.1 1.2 WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 Cells WT1 WT2 WT3 WT4 BS1 BS2 BS3 BS4 0.564 0.958 0.697 0.669 0.019 0.056 0.186 0.011 0.563 0.947 0.924 0.693 <0.001 0.148 0.705 0.647 0.887 0.809 0.034 0.258 <0.001<0.001<0.001<0.001 0.985 0.717 0.023 0.258 <0.001 0.150 0.003 0.263 0.707 0.994 0.992 0.970 0.981 0.636 0.917 0.245 0.395 0.996 0.999 0.920 0.991 0.830 0.997 0.663 G4 on transcribed strand G4 on non-transcribed strand Enrichment Intergenic G4 motifs 0.9 1.0 1.1 1.2 0.386 0.722 0.958 0.978 0.597 0.074 0.101 <0.001
***
*
*
***
***
***
***
***
***
**
a
b
d
c
f
e
h
g
Fig. 3 Bloom syndrome SCE enrichments occur at transcribed G4 motifs. a Intragenic G4 motifs can occur on either the transcribed strand (RNA shown in red), or on the non-transcribed strand;b–h relative SCE enrichments (red points) over random distributions (violin plots) for SCEs overlapping with h for intergenic G4 motifs, and G4 motifs occurring onc intragenic transcribed strands; d transcribed strands; e transcribed strands of active genes; f non-transcribed strands for active genes;g transcribed strands for silent genes; and h non-transcribed strands for silent genes.p-values indicate the fraction of permuted overlaps (out of 1,000 permutations) equal to or higher than overlap with observed SCE regions. Significant p-values are indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001
gene, and performed SCE enrichment analysis for these locations.
Although we found no evidence of SCE enrichments at intergenic
G4 motifs (Fig.
3
b), SCE are enriched at intragenic G4 motifs on
both transcribed and non-transcribed strands (Fig.
3
c, d).
Furthermore, BS-specific SCE enrichments were higher on
transcribed than on non-transcribed strands, this effect is even
strongest for G4 motifs on active transcribed genes (Fig.
3
e, f).
Strikingly, no SCE enrichments were detected for either
0.8 1.2 1.6
Enrichment
G4 motifs non-transcribed strand silent genes 0.8 1.2 1.6
Enrichment
G4 motifs transcribed strand silent genes 0.6 0.8 1.0 1.2 1.4 1.6 Enrichment
G4 motifs transcribed strand active genes
0.6 0.8 1.0 1.2 1.4 1.6 Enrichment
G4 motifs non-transcribed strand active genes
Enrichment 0.6 0.8 1.0 1.2 1.4 1.6 WT
C1.Blm_KO C2.Blm_KO C3.Blm_KO C4.Blm_Het
Cells
WT
C1.Blm_KO C2.Blm_KO C3.Blm_KO C4.Blm_Het
Cells
WT
C1.Blm_KO C2.Blm_KO C3.Blm_KO C4.Blm_Het
Cells
WT
C1.Blm_KO C2.Blm_KO C3.Blm_KO C4.Blm_Het
Cells
WT
C1.Blm_KO C2.Blm_KO C3.Blm_KO C4.Blm_Het
Cells
WT
C1.Blm_KO C2.Blm_KO C3.Blm_KO C4.Blm_Het
Cells
WT
C1.Blm_KO C2.Blm_KO C3.Blm_KO C4.Blm_Het
Cells Genes 0.9 1.0 1.1 Enrichment G4N7 motifs 0.054 <0.001 <0.001 0.002 0.001 0.728 <0.001 0.001 0.003 0.023 0.861 0.001 <0.001 <0.001 0.068 0.889 0.004 <0.001 0.083 0.009 0.734 0.056 0.113 0.050 0.068 0.332 0.038 0.046 0.011 0.412 *** ** *** *** ** ** * *** *** *** *** *** ** ** * * * SCEs/library 80 60 40 20 0 p = 1.4e–27 p = 6.7e–53 p = 2.9e–12 % Matching alleles SCE/G4 100 75 50 25 0 WT
C1.Blm_KOC2.Blm_KOC3.Blm_KOC4.Blm_Het
Discordant cast/EiJ NS p = 0.106 (n=41) p = 0.001 (n=34) p = 0.002 (n=71) p = 0.005 (n=52) NS p = 0.623 (n=10) Discordant 129Sv Concordant cast/EiJ Concordant 129Sv
**
**
**
a
b
c
e
d
f
g
h
transcribed or non-transcribed strand G4 motif in silent genes
(Fig.
3
g, h). These results confirm the synergistic effect of
transcriptional activity and the presence of a G4 motif as a trigger
for SCE formation in BS cells.
SCEs map to G4 motifs in both human and murine
BLM
−/−cells. To confirm that the BS SCE enrichment patterns we
detected in human cells are a direct result of BLM deficiency, we
next generated Blm knockout cells in an F1 hybrid mouse
embryonic stem (ES) cell line (129Sv-Cast/EiJ) by means of the
Crispr/Cas9 technology. We used different combinations of two
guide RNAs to generate loss-of-function mutants by deleting Blm
exon 19, which is critical for Blm’s role in both Holliday junction
resolution
36and G4 unwinding
37. We selected three homozygous
and one heterozygous clones with the desired deletions and
characterized these deletions by Sanger sequencing
(Supplemen-tary Table
2
), measured Blm mRNA expression levels by
quan-titative reverse-transcriptase PCR (qRT-PCR) (Supplementary
Fig.
5
a), and confirmed the elevated SCE rates by Strand-seq
(Fig.
4
a and Supplementary Fig.
5
b, c). Interestingly, we detected
intermediately high SCE rates in the Blm
+/−cells, even though
previous studies reported that cells from heterozygous family
members of BS patients display normal SCE levels
2,38. Similar to
that for the human cells, SCEs in libraries made from the ES cells
could be mapped at kilobase resolution (Supplementary Table
2
).
Using the identified SCE regions, we performed the same
analysis as described above for the human cell lines. As before, we
generated RNA-seq data for each of ES cell clones to assess the
effect of transcriptional activity and G4 strandedness on SCE
enrichments. Although we did not detect any clear increased SCE
enrichments in genes for the Blm mutant cell lines (Fig.
4
b) or an
effect of transcriptional activity (Supplementary Fig 5d, e), we did
confirm that these cells display SCE enrichments at canonical and
alternative G4 motifs (Fig.
4
c and Supplementary Fig.
6
a, b). We
detected significant SCE enrichments at sites of intragenic G4
motifs occurring on both transcribed and non-transcribed strands
in the absence of Blm (Supplementary Fig.
6
c, d) and confirmed
that SCE enrichments in absence of Blm are strongest at G4 motifs
occurring on transcribed strands in active genes (Fig.
4
d–g). As in
the human cell lines, we found no SCE enrichments at sites of
intergenic G4 motifs (Supplementary Fig.
6
e).
The F1 hybrid ES cells we used to generate our Blm mutants
contain over 20 million known heterozygous positions, including
72,660 canonical G4 motifs that only occur on one homolog
(36,547 in the 129 Sv background, and 36,203 in the Cast/EiJ
background). To
find further evidence of a direct link between
G4s and SCEs, we identified all observed SCE regions that overlap
a single discordant G4 motif, and the homologs that these SCEs
occurred on. We found that on average, 69% of informative SCEs
in the Blm
−/−cell lines occurred on the same homolog as the G4
Fig. 4 Confirmation of SCE enrichments at G4 motifs in Blm−/−mouse ES cells.a SCE rates detected WT,Blm−/−, andBlm+/−mouse ES cells. Each grey point represents number of SCEs detected in a single-cell Strand-seq library, red lines indicate mean± SD. p-values were calculated using t-test and ANOVA.b–g Relative SCE enrichments (red points) over random distributions (violin plots) for SCEs overlapping one or more (b) genes; c G4 motifs; and G4 motifs occurring ond transcribed strands of active genes; e transcribed strands for active genes; f transcribed strands for silent genes; and g non-transcribed strands for silent genes.p-values indicate the fraction of permuted overlaps (out of 1,000 permutations) equal to or higher than overlap with observed SCE regions. Significant p-values are indicated as follows: *p < 0.05, **p < 0.01, ***p < 0.001. h Frequency of observed SCE regions occurred on the same homolog as allele-specific G4 motifs. Indicated is the homolog containing G4 motif, concordant indicates SCE occurred on same homolog, discordant indicates SCE occurred on opposite homolog. Number of allelic G4 motifs included in analysis is shown above each bar.p-values were calculated using binomial distributions based on a 50% chance of SCE and G4 motif occurring on the same homolog
100 90 80 70 60 50 40 30 20 10 0
% Libraries with LOH region
Chromosome count
Number of CNVs
Cell line + Passage nr
Cell line + Passage nr
Loss of heterozygosity Aneuploidy
Local copy number variations WT
n=77 n=72 n=77 n=79 n=75 n=83 n=80 n=60 n=79 n=83 n=50 n=76 C1.Blm_KO C2.Blm_KO C3.Blm_KO
p = 0.34 p = 0.073 p = 0.001 p = 0.004 p = 0.13 p = 0.38 P0 P20 P30 P0 P20 P30 P0 P20 P30 P0 P20 P30 P0 P20 P30
Cell line + Passage nr
P0 P20 P30 WT C1.Blm_KO C2.Blm_KO C3.Blm_KO 13qA2 1qB 15qD1 8qC2 None 55 50 45 40 8 6 4 2 0 p < 0.001 p < 0.001 p < 0.001 WT C1.Blm_KO C2.Blm_KO C3.Blm_KO
a
b
c
Fig. 5 Low levels of loss of heterozygosity inBlm−/−mouse ES cells. Frequency ofa unique LOH regions; b aneuploidy; and c local copy number variations detected at in single-cell whole genome sequencing libraries at different passages in WT,Blm−/−, andBlm+/−mouse ES cells. The number of single-cell sequencing libraries included in the analysis is shown above each bar.p-values for LOH events were calculated using binomial distributions, for aneuploidy and CNVs by ANOVA
motifs, which is significantly different (p < 0.01) from the
expected 50% if there was no causal relationship between G4
motifs and SCEs (Fig.
4
h). No significant deviation from the
expected 50/50 ratio was detected in the WT or the Blm
+/−cell
lines. Combined, these results confirm that SCEs mainly form at
G4 structures in absence of Blm and especially at those G4s
present in the transcribed strands of active genes.
LOH is not signi
ficantly increased in Blm
−/−cells. As SCEs are
exchanges of genetic material between identical sister chromatids,
they normally do not result in any mutations. However, if an
exchange event occurs between homologs instead of sister
chro-matids, this can lead to LOH
39. It has previously been shown that
BLM deficient cells display elevated levels of LOH
12–14. However,
these results were obtained using systems that rely on selection of
cells that underwent LOH at a specific locus. Using our F1 ES cell
lines, we could detect and track LOH events throughout the entire
genome based on single-nucleotide polymorphisms between the
parental mouse strains. To do this, we kept the WT and Blm
mutant ES cells in continuous culture for 30 passages (~ 75 cell
divisions), which would result in 3.8 × 10
22offspring cells for each
parental cell, compared to an estimated 1.2 × 10
10cells in an adult
mouse body. We performed single-cell whole-genome sequencing
(scWGS) at different timepoints (passages 0, 20, and 30), and
identified chromosomal regions that underwent LOH (see
Methods). We also identified chromosomal and local copy
number variations (CNVs) to confirm that LOH regions are not
caused by deletions, and to determine if the Blm
−/−cells display
aberrant levels of CNVs.
We did not detect a single LOH region in the WT cells at any
of the three time points, and only four unique LOH regions in the
three Blm
−/−clones (Fig.
5
a). Two of these four regions were
detected in a single library at a single time point, while the two
others were detected at multiple time points and their frequency
increased over time. However, these more frequent LOH regions
occurred on chromosomes 1 and 8, both of which display
increasing levels of trisomy in all four cell lines (Supplementary
Fig.
7
). This suggests that trisomy led to clonal expansion within
the cell populations, and the detected LOH regions had no effect
on cellular proliferation. Although these results do point towards
elevated LOH in Blm
−/−cells, the differences are not significant
and suggest that LOH is an uncommon occurrence, even in the
absence of Blm.
BLM has been linked to chromosome segregation
40and
BLM-deficient cells display a higher frequency of micronuclei
41, both of
which can result in aneuploidy. When we assessed the WT and
Blm
−/−cells for instances of local and chromosomal CNVs, we
No BLM > dHJ resolution MUS81/EME1 or GEN1 No BLM > no G4 unwinding
replication fork stalling
Recombination: bypass G4 lesion by polymerase template switch
Return polymerase to original template strand: dHJ formation G4 formation
BLM deficiency Active transcription
G4 motif on transcribed strand
Active BLM > dHJ dissolution BTRR complex
No sister chromatid exchange Sister chromatid exchange
Fig. 6 BLM helicase suppresses recombination at G4 structures. Model for the role of BLM in suppressing recombination at sites of G4 structures. G4 structures are more likely to form or persist in the absence of unwinding by BLM. They can form at G4 motifs throughout the genome, but formation is promoted by transcription, especially if the G4 motif is present on the transcribed strand. In BLM proficient cells, BLM unwinds the G4 structure before the genomic region is replicated, ensuring smooth DNA replication. In the absence of BLM, G4 structures are not unwound, preventing replication fork progression, and leading to replication fork stalling. Stalled forks require homologous recombination (HR)-mediated repair, leading to formation of a double Holliday junction (dHJ). In the absence of BLM, dHJs cannot be dissolved by the BLM-TOPO3α-RMI1-RMI2 (BTRR) complex and must be resolved by MUS81-EME1 or GEN1, leading to frequent formation of sister chromatid exchanges event, and potentially loss of heterozygosity, and other types of mutations
found that although there are significant differences between the
individual cell lines, no trend can be seen indicating that Blm
−/−cells contain more or fewer such events (Fig.
5
b, c).
Discussion
Elevated SCE rates are a hallmark feature in cells from BS
patients
2,3, but the exact mechanism behind this phenotype is not
fully understood. A major obstacle to unravelling the cause of BS
SCEs was that SCEs cannot be accurately mapped using standard
cytogenetic detection methods. For this study, we used Strand-seq
for SCE detection, as this technique does allow for
high-resolution mapping. Even though the technique is limited by
loss of DNA during preparation of single-cell sequencing
librar-ies, leading to low coverage within individual libraries (~ 1–2%
genome coverage), we show here that SCEs in both normal and
BS cells could be mapped at kilobase resolutions, allowing for
robust analysis on SCE locations and thus their causes.
We show that SCEs frequently occur at sites of G4 structures in
both BLM deficient human and murine cells. While there does
not appear to be a direct effect of transcriptional activity on SCE
enrichment patterns, strong SCE enrichments in BS cells were
observed in transcribed genes containing one or more G4 motifs,
especially when the G4 motif was present on the transcribed
DNA strand. The observation that SCEs were enriched on
homolog-specific G4 motifs in the Blm mutant ES cells provides
further evidence that, in the absence of BLM, G4 structures can
directly trigger SCE formation.
Studies of G4 structures have been hampered by their high
stability, making them resistant against several nucleases
42and
difficult to analyze using standard PCR conditions
34. The use of
Strand-seq bypasses these issues, because SCE regions are
iden-tified as the region between sequencing reads. As such, any SCE
overlap with G4 motifs requires that the G4 motif lies within the
identified SCE region and therefore does not have to be covered
by sequencing reads itself.
Previous studies have shown that BLM is required for
unwinding G4 structures during telomere replication
43, and that
it has a role in regulating expression of genes containing G4
motifs
44,45. Our study is the
first to directly implicate
G4 structures in the increased recombination and genome
instability in cells that lack BLM. These results are consistent with
proposed models of BLM unwinding G4 structures during DNA
replication
37,46and with previous reports that BLM binds and
unwinds G4 structures in vitro
15,16. Our results show that BLM is
required to unwind G4 structures throughout the genome.
G4 structures are known to pose barriers for DNA replication
20and previous studies have shown that specialized helicases such as
Dog-1, Pif1 and FANCJ are required to prevent instability at
G-rich genomic DNA in Caenorhabditis elegans
47, yeast
48, and
man
49. The fact that such other helicases cannot compensate for
loss of BLM suggests that these helicases do not have redundant
functions, but are either specific for subsets of G4 structures, or
that they cooperate to unwind G4 structures, as was proposed for
BLM and FANCJ
50,51. Consistent with this, BLM deficient cells
display elevated levels of G4 structures at telomeres
43, and it
seems logical that this holds true throughout the genome. We
propose that failure to unwind G4 structures in BLM-deficient
cells leads to stalled replication forks, which trigger
recombina-tion and genome instability (Fig.
6
).
SCEs in cells lacking BLM were found to frequently occur in
transcribed genes, supporting that such sites are subject to higher
mutation rates
27. Elevated intragenic mutations rates are likely to
contribute to the strong cancer predisposition associated with BS.
This also helps explain a unique feature of BS, which predisposes
patients to a wide range of cancers instead of towards specific
types of tumors
1. Combined with elevated LOH levels, the
pro-posed chromosome fragility is likely to play a role in the strong
cancer predisposition associated with the syndrome.
Methods
Cell cultures. The following cell lines were obtained from the Corriell Cell Repository: GM07492 and GM07545 (primaryfibroblasts, normal), GM02085 and GM03402 (primaryfibroblasts, BS), GM12891 and GM12892 (EBV-transformed lymphocytes, normal), and GM16375 and GM17361 (EBV-transformed B-lymphocytes, BS). The WT hybrid mouse ES cell line F121.6 (129Sv-Cast/EiJ) was a kind gift from Joost Gribnau (Erasmus University, Rotterdam, The Netherlands).
Fibroblasts were cultured in Dulbecco's modified Eagle's medium (DMEM) (Life Technologies) supplemented with 10% v/v fetal bovine serum (FBS) (Sigma Aldrich) and 1% v/v penicillin–streptomycin (Life Technologies), B-lymphocytes in RPMI1640 (Life Technologies) supplemented with 15% v/v FBS and 1% v/v penicillin–streptomycin.
ES cells were cultured on mitotically arrested mouse embryonicfibroblast cells in DMEM (Life Technologies), supplemented with 15% v/v FBS (Bodinco BV), 1% v/v penicillin–streptomycin, 1% v/v non-essential amino acids (Life Technologies), 50µM 2-mercaptoethanol (ThermoFisher Scientific), and 1,000 U ml−1leukemia
inhibitor factor (Merck). All cells were cultured at 37 °C in 5% CO2. For
Strand-seq, BrdU (Invitrogen) was added to exponentially growing cell cultures at 40µM final concentration. Timing of BrdU pulse was 12 h for ES cells, 18 h for fibroblast cell lines, and 24 h for B-lymphocyte cell lines.
Generation of Blm mutant ES cell lines. Blm mutants were generated using CRISPR/Cas9 genome editing. sgRNAs were designed to cleave the Blm gene at sitesflanking exon 19 and cloned into PX459 plasmid52. Combinations of two plasmids (30μg each) were transfected into F121.6 cells by means of electro-poration (Biorad Genepulser XL). Cells were incubated for 24 h before puromycin (1µg/ml) was added to cell culture medium. After 48 h of selection, resistant colonies were left to grow, picked and expanded. Screening for Blm mutant clones was performed by allele-specific PCR of genomic region containing putative deletion.
qRT-PCR analysis. Exponentially growing cells were collected and RNA was isolated using the Nucleospin RNA kit (Macherey Nagel). Reverse transcription was performed using Superscript II Reverse Transcriptase (Invitrogen) with ran-dom hexamers (Invitrogen). Quantitative PCR was performed using SYBR Green I Master (Roche) on the LightCycler480 (Roche).
Strand-seq and scWGS library preparation. For Strand-seq and WGS, expo-nentially growing cells were collected after BrdU pulse (for Strand-seq) or without any treatment (WGS), and resuspended in nucleo isolation buffer (100 mM Tris-HCl pH 7.4, 150 mM NaCl, 1 mM CaCl2, 0.5 mM MgCl2, 0.1% NP-40, and 2%
bovine serum albumin) supplemented with 10µg ml−1Hoechst 33,258 (Life Technologies) and propidium iodide (Sigma Aldrich). Single nuclei were sorted into 5µl Pro-Freeze-CDM NAO freeze medium (Lonza) + 7.5% dimethyl sulfoxide, in 96-well skirted PCR plates (4Titude), based on low propidium iodide and low Hoechstfluorescence using a MoFlo Astrios cell sorter (Beckman Coulter) or a FACSJazz cell sorter (BD Biosciences). DNA from single cells was processed for Strand-seq23or WGS53. For each experiment, 96 libraries were pooled and
250–450 bp-sized fragments were isolated and purified. DNA quality and con-centrations were assessed using the High Sensitivity dsDNA kit (Agilent) on the Agilent 2100 Bio-Analyzer and on the Qubit 2.0 Fluorometer (Life Technologies). RNA-seq library preparation. Exponentially growing cells were harvested and RNA was isolated using the Nucleospin RNA kit (Macherey Nagel). RNA-sequencing libraries were prepared using the NEBNext Ultra RNA Library Prep kit for Illumina (NEB) combined with the NEBNext rRNA Depletion kit (NEB). Complementary DNA quality and concentrations were assessed using the High Sensitivity dsDNA kit (Agilent) on the Agilent 2100 Bio-Analyzer and on the Qubit 2.0 Fluorometer (Life Technologies).
Illumina sequencing. Clusters were generated on the cBot (HiSeq2500) and single-end 50 bp reads (Strand-seq and RNA-seq) or paired-single-end 150 bp reads (scWGS) were generated were generated using the HiSeq2500 sequencing platform (Illumina).
Bioinformatics
Genome alignment. Indexed bamfiles were aligned to human (GRCh37) or mouse genomes (GRCm38) using Bowtie254for Strand-seq and scWGS libraries, and
STAR aligner55for RNA-seq libraries.
Sister chromatid exchange detection. SCE were identified and mapped with the BAIT software package56, using standard settings. As BAIT also detects stable
> 5% of cells from one cell line were excluded from the analysis. SCEs were assigned to homologs by splitting.bamfiles into separate files for each genetic background based on reads covering informative polymorphisms and using BAIT to identify on which homologs SCEs occurred.
Detection and analysis of SCE hotspots. BAIT-generated.bed files containing the locations of all mapped SCEs were uploaded to the USCS genome browser and hotspots were identified as regions containing multiple overlapping SCEs. p-values were assigned to putative SCE hotspots using a custom R-script based on capture–recapture statistics. Briefly, the genome was divided into bins of the same size as the putative hotspot and the chance offindings the observed number of SCEs in one bin was calculated based on the total number of SCEs detected in the cell line.
Enrichment analysis. A custom Perl script was used for the permutation model. For each of 1,000 permutations, we generated a random number n and shifted all SCEs downstream by n bases on the same chromosome. To prevent small-scale local shifts, we required n to be a random number between 2 and 50 Mbp. If the resulted coordinate exceeded chromosome size we subtracted the size of chromosome, so that the SCE is mapped to beginning part of the chromosome, as if the chromo-some was circular. We also excluded all annotated assembly gaps before our analysis, to prevent permuted SCE mapping to one of the gap regions. We then determined the number of SCEs overlapping with a feature of interest in each permutation, as well as the original SCE regions. All values were normalized to the median permutated value, in order to determine relative SCE enrichments over expected, randomized distributions and to allow for comparison of the different cell lines. Significance was determined based on how many permutations showed the same or exceeding (enrichment) or the same or receding (depletion) overlap with a given genomic feature compared to overlap between the original SCEs and the same feature. Any experimental overlap that lies outside of the 95% confidence interval found in the permutations has a p-value below 0.05 and was deemed significant. Experimental overlaps lying outside of the permuted range were given a p-value below 0.001, as there was a<0.1% (1/1,000) chance of such an overlap occurring by chance. Enrichment analyses for G4 motifs were performed using a 10 Kb SCE region size cutoff, enrichment analysis for genes and promoter regions used a 100 Kb size cutoff, unless specified otherwise. Genome and gene annotations were obtained from Ensembl release 75 (GRCh37 assembly,http://www.ensembl. org). Gene bodies were defined as regions between transcription start sites and
transcription end sites, gene promoters as 1 Kbp regions upstream of transcription start sites. Putative G4 motifs were predicted using custom Perl script by matching genome sequence against following patterns: G3+NxG3+NxG3+NxG3+, where x could
be the ranges of 1–3, 1–7, or 1–12 bp.
RNA-seq analysis. Mapped reads were aligned and quantified using STAR aligner55.
FPKM values were calculated for all genes and based on these genes were assigned active (FPKM> 1) or silent (FPKM < 1) status.
Aneuploidy and CNV detection. Aligned libraries were analyzed as previously described using AneuFinder R package57using the following settings: low-quality alignments (mapping quality score (MAPQ)< 10) and duplicate reads were excluded and read counts in 2 Mb variable-width bins were determined with a 10-state Hidden Markov Model with copy-number 10-states: zero-inflation, null-, mono-, di-, tri-, tetra-, penta-, hexa-, septa-, and octasomy.
LOH detection. Reads were aligned to either 129 Sv or Cast/EiJ genetic background based on covered single-nucleotide polymorphisms (SNPs). Reads lacking infor-mative SNPs were discarded. Reads (129 Sv) were assigned a positive (Crick) orientation, Cast/EiJ reads a negative (Watson) orientation. The resulting.bamfiles were analyzed using BAIT and LOH events were detected as switches from mixed background to pure 129 Sv or Cast/EiJ background in the absence of deletions (as detected using AneuFinder).
Data availability. The Strand-seq, scWGS, and RNA-seq data reported in this paper have been submitted to the Arrayexpress database under accession E-MTAB-5976. SCE enrichment analysis software is available through GitHub (https:// github.com/Vityay/GenomePermute).
Received: 5 October 2017 Accepted: 21 December 2017
References
1. German, J. Bloom syndrome: a mendelian prototype of somatic mutational disease. Med. (Baltim.). 72, 393–406 (1993).
2. Chaganti, R. S., Schonberg, S. & German, J. A manyfold increase in sister chromatid exchanges in Bloom’s syndrome lymphocytes. Proc. Natl Acad. Sci. USA 71, 4508–4512 (1974).
3. van Wietmarschen, N. & Lansdorp, P. M. Bromodeoxyuridine does not contribute to sister chromatid exchange events in normal or Bloom syndrome cells. Nucleic Acids Res. 44, 6787–6793 (2016).
4. Painter, R. B. A replication model for sister-chromatid exchange. Mutat. Res. 70, 337–341 (1980).
5. Wu, L. Role of the BLM helicase in replication fork management. DNA Repair (Amst.). 6, 936–944 (2007).
6. Bradley, M. O., Hsu, I. C. & Harris, C. C. Relationship between sister chromatid exchange and mutagenicity, toxicity and DNA damage. Nature 282, 318–320 (1979).
7. Karow, J. K., Constantinou, A., Li, J. L., West, S. C. & Hickson, I. D. The Bloom’s syndrome gene product promotes branch migration of holliday junctions. Proc. Natl Acad. Sci. USA 97, 6504–6508 (2000).
8. Wu, L. & Hickson, I. D. The Bloom’s syndrome helicase suppresses crossing over during homologous recombination. Nature 426, 870–874 (2003). 9. Davies, S. L., North, P. S. & Hickson, I. D. Role for BLM in replication-fork
restart and suppression of originfiring after replicative stress. Nat. Struct. Mol. Biol. 14, 677–679 (2007).
10. Machwe, A., Xiao, L., Groden, J. & Orren, D. K. The Werner and Bloom syndrome proteins catalyze regression of a model replication fork. Biochemistry 45, 13939–13946 (2006).
11. Rao, V. A. et al. Phosphorylation of BLM, dissociation from topoisomerase IIIalpha, and colocalization with gamma-H2AX after topoisomerase I-induced replication damage. Mol. Cell. Biol. 25, 8925–8937 (2005).
12. LaRocque, J. R. et al. Interhomolog recombination and loss of heterozygosity in wild-type and Bloom syndrome helicase (BLM)-deficient mammalian cells. Proc. Natl Acad. Sci. USA 108, 11971–11976 (2011).
13. Suzuki, T., Yasui, M. & Honma, M. Mutator phenotype and DNA double-strand break repair in BLM helicase-deficient human cells. Mol. Cell. Biol. 36, 2877–2889 (2016).
14. Luo, G. et al. Cancer predisposition caused by elevated mitotic recombination in Bloom mice. Nat. Genet. 26, 424–429 (2000).
15. Sun, H., Karow, J. K., Hickson, I. D. & Maizels, N. The Bloom’s syndrome helicase unwinds G4 DNA. J. Biol. Chem. 273, 27587–27592 (1998). 16. Wu, W. Q., Hou, X. M., Li, M., Dou, S. X. & Xi, X. G. BLM unfolds
G-quadruplexes in different structural environments through different mechanisms. Nucleic Acids Res. 43, 4614–4626 (2015).
17. Huber, M. D., Duquette, M. L., Shiels, J. C. & Maizels, N. A conserved G4 DNA binding domain in RecQ family helicases. J. Mol. Biol. 358, 1071–1080 (2006). 18. Bochman, M. L., Paeschke, K. & Zakian, V. A. DNA secondary structures:
stability and function of G-quadruplex structures. Nat. Rev. Genet. 13, 770–780 (2012).
19. Rhodes, D. & Lipps, H. J. G-quadruplexes and their regulatory roles in biology. Nucleic Acids Res. 43, 8627–8637 (2015).
20. Lopes, J. et al. G-quadruplex-induced instability during leading-strand replication. EMBO J. 30, 4033–4046 (2011).
21. Aguilera, A. & Gomez-Gonzalez, B. Genome instability: a mechanistic view of its causes and consequences. Nat. Rev. Genet. 9, 204–217 (2008).
22. Falconer, E. et al. DNA template strand sequencing of single-cells maps genomic rearrangements at high resolution. Nat. Methods 9, 1107–1112 (2012). 23. Sanders, A. D., Falconer, E., Hills, M., Spierings, D. C. J. & Lansdorp, P. M.
Single-cell template strand sequencing by Strand-seq enables the
characterization of individual homologs. Nat. Protoc. 12, 1151–1176 (2017). 24. Durkin, S. G. & Glover, T. W. Chromosome fragile sites. Annu. Rev. Genet. 41,
169–192 (2007).
25. Glover, T. W. & Stein, C. K. Induction of sister chromatid exchanges at common fragile sites. Am. J. Hum. Genet. 41, 882–890 (1987).
26. Sollier, J. & Cimprich, K. A. Breaking bad: R-loops and genome integrity. Trends Cell. Biol. 25, 514–522 (2015).
27. Aguilera, A. & Gaillard, H. Transcription and recombination: when RNA meetsDNA. Cold Spring Harb. Perspect. Biol. 6,https://doi.org/10.1101/ cshperspect.a016543(2014).
28. Grierson, P. M., Acharya, S. & Groden, J. Collaborating functions of BLM and DNA topoisomerase I in regulating human rDNA transcription. Mutat. Res. 743-744, 89–96 (2013).
29. Chang, E. Y. et al. RECQ-like helicases Sgs1 and BLM regulate R-loop-associated genome instability. J. Cell Biol.,https://doi.org/10.1083/ jcb.201703168(2017).
30. Huppert, J. L. & Balasubramanian, S. G-quadruplexes in promoters throughout the human genome. Nucleic Acids Res. 35, 406–413 (2007).
31. Eddy, J. & Maizels, N. Conserved elements with potential to form polymorphic G-quadruplex structures in thefirst intron of human genes. Nucleic Acids Res. 36, 1321–1333 (2008).
32. Todd, A. K., Johnston, M. & Neidle, S. Highly prevalent putative quadruplex sequence motifs in human DNA. Nucleic Acids Res. 33, 2901–2907 (2005).
33. Hansel-Hertsch, R. et al. G-quadruplex structures mark human regulatory chromatin. Nat. Genet. 48, 1267–1272 (2016).
34. Chambers, V. S. et al. High-throughput sequencing of DNA G-quadruplex structures in the human genome. Nat. Biotechnol. 33, 877–881 (2015). 35. Maizels, N. G4 motifs in human genes. Ann. N. Y. Acad. Sci. 1267, 53–60
(2012).
36. Kim, Y. M. & Choi, B. S. Structure and function of the regulatory HRDC domain from human Bloom syndrome protein. Nucleic Acids Res. 38, 7764–7777 (2010).
37. Chatterjee, S. et al. Mechanistic insight into the interaction of BLM helicase with intra-strand G-quadruplex structures. Nat. Commun. 5, 5556 (2014). 38. Ellis, N. A. et al. Somatic intragenic recombination within the mutated locus
BLM can correct the high sister-chromatid exchange phenotype of Bloom syndrome cells. Am. J. Hum. Genet. 57, 1019–1027 (1995).
39. Moynahan, M. E. & Jasin, M. Loss of heterozygosity induced by a chromosomal double-strand break. Proc. Natl Acad. Sci. USA 94, 8988–8993 (1997). 40. Chan, K. L., North, P. S. & Hickson, I. D. BLM is required for faithful
chromosome segregation and its localization defines a class of ultrafine anaphase bridges. EMBO J. 26, 3397–3409 (2007).
41. Yankiwski, V., Marciniak, R. A., Guarente, L. & Neff, N. F. Nuclear structure in normal and Bloom syndrome cells. Proc. Natl Acad. Sci. USA 97, 5214–5219 (2000).
42. Bishop, J. S. et al. Intramolecular G-quartet motifs confer nuclease resistance to a potent anti-HIV oligonucleotide. J. Biol. Chem. 271, 5698–5703 (1996). 43. Drosopoulos, W. C., Kosiyatrakul, S. T. & Schildkraut, C. L. BLM helicase
facilitates telomere replication during leading strand synthesis of telomeres. J. Cell. Biol. 210, 191–208 (2015).
44. Smestad, J. A. & Maher, L. J. III. Relationships between putative G-quadruplex-forming sequences, RecQ helicases, and transcription. BMC Med. Genet. 16, 91 (2015).
45. Nguyen, G. H. et al. Regulation of gene expression by the BLM helicase correlates with the presence of G-quadruplex DNA motifs. Proc. Natl Acad. Sci. USA 111, 9905–9910 (2014).
46. Croteau, D. L., Popuri, V., Opresko, P. L. & Bohr, V. A. Human RecQ helicases in DNA repair, recombination, and replication. Annu. Rev. Biochem. 83, 519–552 (2014).
47. Cheung, I., Schertzer, M., Rose, A. & Lansdorp, P. M. Disruption of dog-1 in Caenorhabditis elegans triggers deletions upstream of guanine-rich DNA. Nat. Genet. 31, 405–409 (2002).
48. Paeschke, K., Capra, J. A. & Zakian, V. A. DNA replication through G-quadruplex motifs is promoted by the Saccharomyces cerevisiae Pif1 DNA helicase. Cell 145, 678–691 (2011).
49. Castillo Bosch, P. et al. FANCJ promotes DNA synthesis through G-quadruplex structures. EMBO J. 33, 2521–2533 (2014).
50. Suhasini, A. N. et al. Interaction between the helicases genetically linked to Fanconi anemia group J and Bloom’s syndrome. EMBO J. 30, 692–705 (2011). 51. Sarkies, P. et al. FANCJ coordinates two pathways that maintain epigenetic
stability at G-quadruplex DNA. Nucleic Acids Res. 40, 1485–1498 (2012). 52. Ran, F. A. et al. Genome engineering using the CRISPR-Cas9 system. Nat.
Protoc. 8, 2281–2308 (2013).
53. van den Bos, H. et al. Single-cell whole genome sequencing reveals no evidence for common aneuploidy in normal and Alzheimer’s disease neurons. Genome Biol. 17, 116 (2016).
54. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods 9, 357–359 (2012).
55. Dobin, A. et al. STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21 (2013).
56. Hills, M. et al. BAIT: Organizing genomes and mapping rearrangements in single cells. Genome Med. 5, 82 (2013).
57. Bakker, B. et al. Single-cell sequencing reveals karyotype heterogeneity in murine and human malignancies. Genome Biol. 17, 115 (2016).
Acknowledgements
We thank Dirk Hockemeyer, Marcel van Vugt, and Peter Stirling for critical reading of this manuscript, Inge Kazemier and Karina Hoekstra-Wakker for technical assistance, and Ester Falconer, Mark Hills, and all members of the Lansdorp laboratories in Van-couver and Groningen for discussions and feedback. Financial support was provided by an Advanced Grant from the European Research Council to P.M.L.
Author contributions
N.v.W. and P.M.L. conceived and designed the study. N.v.W. and S.M. created and characterized Blm mutant cell lines. N.v.W. and N.H. performed Strand-seq, scWGS, and RNA-seq experiments. D.C.S.J. supervised next-generation sequencing efforts. N.v.W. and V.G. analysed sequencing data. N.v.W. wrote the manuscript with assistance from S.M., P.M.L. and all authors. P.M.L. supervised the project.
Additional information
Supplementary Informationaccompanies this paper at https://doi.org/10.1038/s41467-017-02760-1.
Competing interests:The authors declare no competingfinancial interests. Reprints and permissioninformation is available online athttp://npg.nature.com/ reprintsandpermissions/
Publisher's note:Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visithttp://creativecommons.org/ licenses/by/4.0/.